Developing an optimised remote sensing and GIS methodology for monitoring urban growth in major towns in North West Province, South Africa

Abstract

In the North West Province of South Africa, as well as in South Africa in general, there is a rapid rate of urban expansion. The rapid rate requires techniques that can effectively detect and quantify the rate of change. The aim of the study was to establish an optimised Remote Sensing and GIS methodology for monitoring urban growth in major towns in the North West Province. The study was conducted in the province's four main towns, namely Mahikeng, Rustenburg, Klerksdorp and Potchefstroom. SPOT images of the towns from dry season dates in 1999, 2013 and 2017 were acquired from the South African Space Agency (SANSA). Processing of the images to map multi-temporal urban sprawl established a methodology for monitoring urban sprawl in the province, given the unique characteristics of the urban metrics. The images were processed using the combination of ERDAS Imagine 2016 and eCognition Developer 9 software. The recommended methodology consists of establishing similar image data units if there are differences in radiometric resolution, as well as common pixel size and projection. Following a comparison with the widely used, parametric and accurate Maximum Likelihood Classifier (MLC), the non-parametric approach of Random Forest (RF) classification was established as more accurate in discriminating the urban land in the four towns studied. The RF classifier is an Object Based Image Analysis (OBIA) approach. The generation of objects was optimised using the multi-resolution segmentation algorithm. Segmentation level 3 at the scale parameter of 160 gave ideal segmentation results from the multi-resolution segmentation. Boolean GIS overlay analysis using the built up area class from each image then enabled mapping and quantification of the urban sprawl. It is concluded that the non-parametric classification approach using a non-parametric classifier like Random Forest is more accurate for delineating the urban metrics of the towns in the North West Province of South Africa, whose urban land is characterised by a mixture of formal built up areas andthe scantly built informal settlements with gravel roads, bare land and small dwellings ('shacks') made from iron roofing sheets. The methodology is recommended for use in spatial assessments of urban growth in the towns of the North West Province of South Africa as well as other towns in the country or in other countries with similar urban metrics.